Website fingerprinting on Nym: Attacks and Defenses

Website Fingerprinting on Nym: Attacks and Defenses This directory contains the datasets for the paper "Website Fingerprinting on Nym: Attacks and Defenses" accepted at Privacy Enhancing Technologies Symposium (PETs) 2026. The datasets include Website Fingerprinting (WF) experiments with Tor and Nym mixnet configurations. Directory Structure full_list/ Complete dataset of monitored and unmonitored websites captured across different network configurations for websites in the Full List. Format: <website_url> <num_packets> <timestamp1>:<size1> <timestamp2>:<size2> ... where positive sizes are incoming packets and negative sizes are outgoing packets. reduced_list/ Processed traffic data with various Nym mixnet configurations and defense mechanisms for websites in the Reduced List. Each subdirectory contains aggregated traffic traces in time-series format: <website_url> <num_packets> <timestamp1>:<size1> <timestamp2>:<size2> ... where positive sizes are incoming packets and negative sizes are outgoing packets. traffic_captures/ Same traces as in `reduced_list/` but stored as individual trace files. Format: <timestamp>\t<signed_volume> (one packet per line), where timestamp is in <website_index>-<page_number>. train_test_WF/ Pre-processed pickle files ready for Website Fingerprinting attack training using the ExplainWF framework. Generated from `traffic_captures/` using the transform_to_ml.ipynb notebook. Each pickle file contains a tuple (X_train, X_test, y_train, y_test) with 60/40 train/test split and stratification. overheads/ Processed data for overhead analysis including: Latency measurements Bandwidth overhead calculations Traffic volume comparisons Serialized analysis results (`infos.pkl`) Configuration Details This section describes all network configurations available across the different dataset directories. Baseline Configurations No Anonymization: no_proxy - Direct connection without any anonymization network Nym Network Variants: nym_labnet_open - Nym lab network (open world) nym_mainet_open - Nym mainnet (open world) nym_labnet_full_closed - Nym lab network (full dataset, closed world) nym_labnet_full_open - Nym lab network (full dataset, open world) nym_labnet_reduced_closed - Nym lab network (reduced dataset, closed world) nym_labnet_reduced_open - Nym lab network (reduced dataset, open world) nym_mainnet_full_closed - Nym mainnet (full dataset, closed world) nym_mainnet_full_open - Nym mainnet (full dataset, open world) nym_mainnet_reduced_closed - Nym mainnet (reduced dataset, closed world) nym_mainnet_reduced_open - Nym mainnet (reduced dataset, open world) Tor Network Variants: tor_open - Tor network (open world) tor_full_closed - Tor network (full dataset, closed world) tor_full_open - Tor network (full dataset, open world) tor_reduced_closed - Tor network (reduced dataset, closed world) tor_reduced_open - Tor network (reduced dataset, open world) Fair Evaluation: data_nym_fair_open - Nym network (fair evaluation setup) data_tor_fair_open - Tor network (fair evaluation setup) Nym Mixnet Parameter Configurations These configurations test different Nym mixnet parameter settings: configuration00_default - Default Nym configuration (baseline) configuration01_lqp10 - Proxy-side Poisson rate = 10 packets/s (LQP=10) configuration02_lqp40 - Proxy-side Poisson rate = 40 packets/s (LQP=40) configuration03_ll0 - No loop cover traffic (LL=0) configuration04_lqp0_ll0 - No proxy-side cover traffic, no loops (LQP=0, LL=0) configuration05_lqp0_lm100_ll0 - No proxy cover, no loops, mix delay = 100ms (LQP=0, LM=100, LL=0) configuration06_lqp0_lm0_ll0 - No proxy cover, no loops, no mix delay (LQP=0, LM=0, LL=0) configuration07_lm0 - No mix delay (LM=0) configuration08_lm20 - Mix delay = 20ms (LM=20) configuration09_lm100 - Mix delay = 100ms (LM=100) configuration10_lqr10 - Network requester Poisson rate = 10 packets/s (LQR=10) configuration11_lqr20 - Network requester Poisson rate = 20 packets/s (LQR=20) configuration12_lqp0_ll0_lqr10 - No proxy cover, no loops, requester rate = 10 packets/s (LQP=0, LL=0, LQR=10) Defense Mechanisms Nym Defenses: nym_defence_front - Nym with FRONT defense nym_defence_poissoff - Nym with Poisson-based defense nym_defence_wtf_pad - Nym with WTF-PAD defense nym_defence_wtf4nym - Nym with WTF4NYM defense (custom WTF-PAD variant for Nym) Tor Defenses: tor_defence_front - Tor with FRONT defense tor_defence_tamaraw - Tor with Tamaraw defense tor_defence_wtf-pad - Tor with WTF-PAD defense Flow Correlation Analysis correlation_with_defense - Traffic correlation analysis with defenses enabled correlation_without_defense - Traffic correlation analysis without defenses《Nym网站指纹分析：攻击与防御》 本数据集目录源自2026年隐私增强技术研讨会（Privacy Enhancing Technologies Symposium, PETs）收录的同名论文，包含针对Tor与Nym混合网络（mixnet）配置的网站指纹（Website Fingerprinting, WF）实验数据集。 ## 目录结构 ### `full_list/` 完整数据集，涵盖全列表网站在多种网络配置下的受监控与未受监控网站流量。格式为：`<网站URL> <数据包总数> <时间戳1>:<数据包大小1> <时间戳2>:<数据包大小2> ...`，其中正数大小代表入站数据包，负数大小代表出站数据包。 ### `reduced_list/` 经预处理的流量数据，对应精简列表网站的各类Nym混合网络配置与防御机制。每个子目录包含聚合后的时序格式流量轨迹，格式为：`<网站URL> <数据包总数> <时间戳1>:<数据包大小1> <时间戳2>:<数据包大小2> ...`，其中正数大小代表入站数据包，负数大小代表出站数据包。 ### `traffic_captures/` 与`reduced_list/`内的流量轨迹完全一致，但存储为独立轨迹文件。格式为：`<时间戳> <带符号的流量体积>`，每行对应一个数据包，其中时间戳格式为`<网站索引>-<页面编号>`。 ### `train_test_WF/` 可直接用于网站指纹攻击训练的预处理 pickle 文件，通过`transform_to_ml.ipynb`脚本从`traffic_captures/`生成，适配ExplainWF框架。每个pickle文件包含元组`(X_train, X_test, y_train, y_test)`，采用60/40的训练/测试划分并保持分层采样。 ### `overheads/` 用于开销分析的处理后数据，包括：延迟测量结果、带宽开销计算数据、流量体积对比数据，以及序列化分析结果文件`infos.pkl`。 ## 配置详情 本小节介绍各数据集目录中使用的全部网络配置。 ### 基准配置 无匿名化： - `no_proxy`：无代理直连，未使用任何匿名网络 ### Nym网络变体 - `nym_labnet_open`：Nym实验室网络（开放世界场景） - `nym_mainnet_open`：Nym主网（开放世界场景） - `nym_labnet_full_closed`：Nym实验室网络（全数据集，封闭世界场景） - `nym_labnet_full_open`：Nym实验室网络（全数据集，开放世界场景） - `nym_labnet_reduced_closed`：Nym实验室网络（精简数据集，封闭世界场景） - `nym_labnet_reduced_open`：Nym实验室网络（精简数据集，开放世界场景） - `nym_mainnet_full_closed`：Nym主网（全数据集，封闭世界场景） - `nym_mainnet_full_open`：Nym主网（全数据集，开放世界场景） - `nym_mainnet_reduced_closed`：Nym主网（精简数据集，封闭世界场景） - `nym_mainnet_reduced_open`：Nym主网（精简数据集，开放世界场景） ### Tor网络变体 - `tor_open`：Tor网络（开放世界场景） - `tor_full_closed`：Tor网络（全数据集，封闭世界场景） - `tor_full_open`：Tor网络（全数据集，开放世界场景） - `tor_reduced_closed`：Tor网络（精简数据集，封闭世界场景） - `tor_reduced_open`：Tor网络（精简数据集，开放世界场景） ### 公平评估配置 - `data_nym_fair_open`：Nym网络（公平评估设置） - `data_tor_fair_open`：Tor网络（公平评估设置） ### Nym混合网络参数配置 本系列配置用于测试不同的Nym混合网络参数设置： - `configuration00_default`：默认Nym配置（基准基线） - `configuration01_lqp10`：代理端泊松速率=10数据包/秒（LQP=10） - `configuration02_lqp40`：代理端泊松速率=40数据包/秒（LQP=40） - `configuration03_ll0`：无循环覆盖流量（LL=0） - `configuration04_lqp0_ll0`：无代理端覆盖流量、无循环覆盖流量（LQP=0, LL=0） - `configuration05_lqp0_lm100_ll0`：无代理端覆盖流量、无循环覆盖流量，混合延迟=100ms（LQP=0, LM=100, LL=0） - `configuration06_lqp0_lm0_ll0`：无代理端覆盖流量、无循环覆盖流量，无混合延迟（LQP=0, LM=0, LL=0） - `configuration07_lm0`：无混合延迟（LM=0） - `configuration08_lm20`：混合延迟=20ms（LM=20） - `configuration09_lm100`：混合延迟=100ms（LM=100） - `configuration10_lqr10`：网络请求端泊松速率=10数据包/秒（LQR=10） - `configuration11_lqr20`：网络请求端泊松速率=20数据包/秒（LQR=20） - `configuration12_lqp0_ll0_lqr10`：无代理端覆盖流量、无循环覆盖流量，请求端速率=10数据包/秒（LQP=0, LL=0, LQR=10） ### 防御机制 #### Nym防御方案 - `nym_defence_front`：搭载FRONT防御的Nym网络 - `nym_defence_poissoff`：搭载泊松型防御的Nym网络 - `nym_defence_wtf_pad`：搭载WTF-PAD防御的Nym网络 - `nym_defence_wtf4nym`：搭载WTF4NYM防御的Nym网络（针对Nym定制的WTF-PAD变体） #### Tor防御方案 - `tor_defence_front`：搭载FRONT防御的Tor网络 - `tor_defence_tamaraw`：搭载Tamaraw防御的Tor网络 - `tor_defence_wtf-pad`：搭载WTF-PAD防御的Tor网络 ## 流量相关性分析 - `correlation_with_defense`：启用防御机制时的流量相关性分析 - `correlation_without_defense`：未启用防御机制时的流量相关性分析